How Mantle Chat fits together

Mantle brings team communication and AI work into the same workspace. This page helps you choose the right place for each kind of work.

​

Choose where to work

​

Chats, channels, and direct messages

These three places can all contain messages, but they begin with different people in mind:

• Chat: starts with an AI model or agent. Use it to ask questions, draft, research, or analyze files. Teammates with access to the workspace can still open the chat and continue it.
• Channel: starts with the team. Use it for project updates, decisions, files, and discussion. Mention an AI model or agent with @ when you want AI help inside that discussion.
• Direct message: starts with one teammate. Use it for a private one-to-one conversation inside the organization.

If the main participant is AI, choose a chat. If the main participants are a group of people, choose a channel. If the conversation is only between you and one teammate, choose a direct message. Select the + beside the workspace selector to create a new chat, channel, canvas, or direct message. Agents and Tasks have their own + buttons in the left rail.

​

AI work can be shared

An AI chat does not have to end with one person. In a team workspace, other members can open the conversation, see the context, and ask a follow-up question. You can also mention a model or agent inside a channel, so its answer becomes part of the team discussion. This is useful when a decision needs both AI help and human review. Everyone sees how the answer was reached, not just the final copy.

​

Knowledge stays close to the conversation

Mantle has two main kinds of shared knowledge:

• Canvases are living documents that people edit together.
• Workspace context contains instructions, files, and selected canvases that AI can use across the workspace.

Agents can have their own smaller set of knowledge for a specific job.

​

Agents do repeatable work

An agent combines a model with a job description, knowledge, and optional connected tools. A task tells that agent what to do, when to run, and which channel should receive the result. For example, a support agent could read new feedback from a connected app, group it by theme, and post a morning digest in #customer-feedback.